CN103809937A - Intervisibility parallel processing method based on GPU - Google Patents

Abstract

The invention discloses an intervisibility parallel processing method based on a GPU. The intervisibility parallel processing method based on the GPU comprises the following steps: step 1, a GPU programming environment is created; step 2, a sight line connected with an observation point and a target point is calculated; step 3, data are written into the GPU; step 4, the intervisibility conditions of all points on each segment of the sight line are calculated in parallel; step 6, synchronization points are set; step 6, an intervisibility result is judged in parallel; step 7, the data are read out to a CPU. According to the intervisibility parallel processing method based on the GPU, a CPU and GPU heterogeneous mixed system structure can be supported, and efficient and parallel running of intervisibility calculation is realized through effective utilization of optimization resources such as a novel processor, a novel communicator and a novel synchronizer; by means of organically combing a CUDA structure with an intervisibility calculation method and fully utilizing the CUDA parallel storage and communication mechanism, a layered and parallel high-efficiency intervisibility method is realized, intervisibility calculation accuracy is guaranteed, and calculation time is effectively shortened at the same time.

Description

A kind of intervisibility method for parallel processing based on GPU

Technical field

The present invention relates to intervisibility method for parallel processing technical field, particularly a kind of based on GPU(Graphic Processing Unit, Graphics Processing Unit) intervisibility method for parallel processing.

Background technology

Intervisibility processing is in three dimensions, to calculate the intervisibility situation on line between any given 2, is requisite part in analogue system.Existing intervisibility disposal route take Grid DEM landform as basis form method more, for example Janus algorithm, Dyntacs algorithm, ModSAF algorithm and Bresenham algorithm, the principle of these intervisibility disposal routes is basically identical, difference is elevation interpolating method and intervisibility judgment principle, makes the precision of intervisibility processing and efficiency different.The intervisibility computing method of above-mentioned serial can not solve accuracy and two problems of rapidity that intervisibility calculates simultaneously.

Existing intervisibility is processed serial execution on CPU mostly, and its algorithm focuses on improving the efficiency of single sight line intervisibility computing method, but because intervisibility computation complexity is still O (N2), so efficiency improves and is not obvious.

Aspect intervisibility parallel processing, the people such as Ware adopt some Region Segmentation strategies on computer cluster, to realize parallel intervisibility and calculate; The people such as Kidner have designed a kind of triangle irregular network of multiple dimensioned implicit expression to support the intervisibility under multiple resolution to calculate; The people such as Mineter realize by set up complete intervisibility database in the distributed system of high-throughput parallel that intervisibility calculates.Above-mentioned intervisibility method for parallel processing all mainly focuses in distributed system and realizes parallel intervisibility processing, carries out the communication of intervisibility processing with synchronous by network.Therefore, the limited efficiency of above-mentioned intervisibility method for parallel processing is in the environmental baseline of distributed system.

Summary of the invention

The object of the invention is to provide a kind of intervisibility parallel calculating method based on GPU, solves the not high problem of intervisibility counting yield in existing analogue system, in guaranteeing computational accuracy, effectively reduces the time that intervisibility calculates.

Intervisibility method for parallel processing based on GPU provided by the invention comprises the steps:

S1: build GPU programmed environment;

S2: calculate the sight line that connects observation point and impact point;

S3: write data to GPU;

S4: every some intervisibility situation on parallel computation sight line line segment;

S5: synchronous point is set;

S6: walk abreast and judge intervisibility result;

S7: sense data is to CPU.

Preferably, described GPU programmed environment comprises hardware environment and software environment, and wherein hardware environment comprises CPU, supports the display chip of CUDA framework and connects the pci bus of CPU and display chip; Software environment comprises C/C++ compiler and CUDA.

Preferably, described step S2 comprises following sub-step:

S2.1: read in the position of observation point, position, the terrain data of observation point and the terrain data of impact point of impact point;

S2.2: determine by the sight line line segment of observation point and impact point on CPU;

S2.3: the position of observation point and impact point is converted to geocentric coordinate system from geographic coordinate system;

S2.4: calculate the sight line line segment between observation point and impact point under geocentric coordinate system.

Preferably, the terrain data of observation point and impact point respectively comprises longitude, latitude and height.

Preferably, described step S3 comprises following sub-step:

S3.1: the video memory that all data is written to GPU from the internal memory of CPU;

S3.2: large and that remain unchanged data volume terrain data is put into texture cache from video memory and accelerate to read;

S3.3: constant buffer memory is put into in frequent observation station of accessing and impact point position in calculating and accelerate to read.

Preferably, described step S4 comprises following sub-step:

In S4.1:GPU, a corresponding sight line line segment of thread block, distributes shared drive to preserve communication data;

S4.2: the thread mean allocation in each thread block is calculated the part discrete point in sight line;

S4.3: all threads are carried out identical discrete point intervisibility judgement simultaneously.

Preferably, described step S4.3 is: utilize interpolation computing method to be calculated the geocentric coordinate of corresponding discrete point by the geocentric coordinate of observation point and impact point, calculated longitude, latitude and the height of this discrete point by the geocentric coordinate of this discrete point, obtain the height of the landform of this discrete point position according to the longitude of this discrete point and latitude inquiry terrain data, by relatively judging whether the height of this discrete point is greater than the height of the landform of this discrete point position, and judged result is write to shared drive.

Preferably, described step S5 is: the thread in GPU is arranged to synchronous point, judge until the interior all threads of same thread block all complete every some intervisibility on sight line line segment the calculating that just continues next step.

Preferably, described step S6 is: the each thread block in GPU is by the intervisibility judged result in parallel traversal method shared drive, if the height of all discrete points is all greater than the height of the landform of this position in sight line, judge this sight line intervisibility, otherwise judge not intervisibility of this sight line, and the intervisibility result of this sight line is kept on the video memory of GPU.

Preferably, described step S7 is: the intervisibility judged result of all sight lines that GPU is obtained writes back CPU internal memory, copies to CPU internal memory from the video memory of GPU by pci bus.

The present invention has following beneficial effect:

Compared with the intervisibility method for parallel processing of prior art, intervisibility method for parallel processing of the present invention can be supported CPU, GPU heterogeneous mix architecture, effectively utilizes new types of processors, communication, synchronous etc. optimizes resource, realizes the efficient parallel operation that intervisibility calculates; By CUDA framework and intervisibility computing method are organically combined, make full use of the parallel storage of CUDA and communication mechanism, realize the high-effect intervisibility method of hierarchic parallel, when guaranteeing intervisibility computational accuracy, effectively reduce computing time.

Accompanying drawing explanation

The process flow diagram of the intervisibility method for parallel processing based on GPU that Fig. 1 provides for the embodiment of the present invention;

Fig. 2 is that the exchanges data of the intervisibility parallel calculating method based on GPU is moved towards schematic diagram;

Fig. 3 is tree-shaped addition schematic diagram.

Embodiment

Below in conjunction with drawings and Examples, summary of the invention of the present invention is further described.

As shown in Figure 1, the intervisibility method for parallel processing based on GPU that the present embodiment provides comprises the steps:

S1: build GPU programmed environment;

S2: calculate the sight line that connects observation point and impact point;

S3: write data to GPU;

S4: every some intervisibility situation on parallel computation sight line line segment;

S5: synchronous point is set;

S6: walk abreast and judge intervisibility result;

S7: sense data is to CPU.

In above-mentioned steps S1, GPU programmed environment comprises hardware environment and software environment, wherein hardware environment comprises CPU, supports the display chip of CUDA framework and connects the pci bus of CPU and display chip, and using CPU as main frame (host), using GPU as equipment (device); Software environment comprises C/C++ compiler and CUDA.

Above-mentioned steps S2 comprises following sub-step:

S2.1: read in the position of observation point, position, the terrain data of observation point and the terrain data of impact point of impact point; In the present embodiment, the terrain data of observation point and impact point respectively comprises longitude, latitude and height;

S2.2: determine by the sight line line segment of observation point and impact point on CPU;

S2.3: the position of observation point and impact point is converted to geocentric coordinate system from geographic coordinate system;

S2.4: calculate the sight line line segment between observation point and impact point under geocentric coordinate system.

The conversion relational expression that the position of observation point and impact point is converted to geocentric coordinate system from geographic coordinate system is:

X=(N+H) cos (B) cos (L) formula (1)

Y=(N+H) cos (B) sin (L) formula (2)

Z=[N (1-e_2_C)+H] sin (B) formula (3)

In formula (1), formula (2) and formula (3), (X, Y, Z) is the geocentric coordinate of observation point or impact point; L is observation point or the impact point longitude in geographic coordinate system; B is observation point or the impact point latitude in geographic coordinate system; H is observation point or the impact point height in geographic coordinate system; E_2_C is ellipsoid the first excentricity square, and

$e\_2\_C=\frac{a^2-b^2}{a^2}$ Formula (4)

In formula (4), a is semimajor axis of ellipsoid; B is semiminor axis of ellipsoid; In the present embodiment, a=6378137.0; And b=6356752.3142.

In formula (1), formula (2) and formula (3), N is radius of curvature in prime vertical, and

$N=\frac{a}{\sqrt{1-e\_2\_C*{\left[\sin \left(B\right)\right]}^2}}$ Formula (5)

Above-mentioned steps S3 is: the sight line line segment calculating on CPU and terrain data are copied on GPU, particularly, the storage space that can store sight line line segment and terrain data is set on the video memory of GPU, by pci bus, sight line line segment and terrain data is transferred to the storage space of the video memory of GPU from the internal memory of CPU.

Above-mentioned steps S3 comprises following sub-step:

S3.1: all data are written to the video memory of GPU from the internal memory of CPU, as shown in the label 1 in Fig. 2;

S3.2: large and that remain unchanged data volume terrain data is put into texture cache from video memory and accelerate to read, as shown in the label 2 in figure Fig. 2;

S3.3: constant buffer memory is put into in frequent observation station of accessing and impact point position in calculating and accelerate to read, as shown in the label 3 in figure Fig. 2.

Above-mentioned steps S4 comprises following sub-step:

In S4.1:GPU, a corresponding sight line line segment of thread block, distributes shared drive to preserve communication data;

S4.2: the thread mean allocation in each thread block is calculated the part discrete point in sight line;

S4.3: all threads are carried out identical discrete point intervisibility judgement simultaneously.

Above-mentioned steps S4.3 is: utilize interpolation computing method to be calculated the geocentric coordinate of corresponding discrete point by the geocentric coordinate of observation point and impact point, calculated longitude, latitude and the height of this discrete point by the geocentric coordinate of this discrete point, obtain the height of the landform of this discrete point position according to the longitude of this discrete point and latitude inquiry terrain data, by relatively judging whether the height of this discrete point is greater than the height of the landform of this discrete point position, and judged result is write to shared drive.

The conversion relational expression that the position of observation point and impact point is converted to geographic coordinate system from geocentric coordinate system is:

$L=\left\{\begin{array}{c}\arctan \left(\frac{Y}{X}\right),X>0\\ \arctan \left(\frac{Y}{X}\right)+\mathrm{\&pi;},X<0,Y>0\\ \arctan \left(\frac{Y}{X}\right)-\mathrm{\&pi;},X<0,Y<0\\ \frac{\mathrm{\&pi;}}{2},X=0,Y>0\\ -\frac{\mathrm{\&pi;}}{2},X=0,Y<0\end{array}\right\}$ Formula (6)

$B=\arctan \left[\frac{Z+b*e\_2\mathrm{nd}\_2\_C*\sin {\left(\mathrm{\&mu;}\right)}^3}{p-a*e\_2\_C*\cos {\left(\mathrm{\&mu;}\right)}^3}\right];$ Formula (7)

$H=p*\cos \left(B\right)+Z*\sin \left(B\right)-a\sqrt{1-e\_2\_C*\sin {\left(B\right)}^2};$ Formula (8)

In formula (6), formula (7) and formula (8), e_2nd_2_C is ellipsoid the second excentricity square, and

$e\_2\mathrm{nd}\_2\_C=\frac{a^2-b^2}{b^2};$ Formula (9)

$p=\sqrt{X^2+Y^2};$ Formula (10)

$\mathrm{\&mu;}=\arctan \left(\frac{a*Z}{b*p}\right).$ Formula (11)

Above-mentioned interpolation computing method is:

The computing formula of utilizing distance weighted method to calculate the height of interpolated point is:

$z=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\left(\frac{z_i}{d_i^2}\right)}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{d_i^2}};$ Formula (12)

In formula (12), n=4; z _ifor the height of graticule mesh node; d _ifor graticule mesh node is to the distance of interpolated point.The height of putting on graticule mesh limit adopts simple linear interpolation to calculate.Known two adjacent graticule mesh nodes are respectively A (x ₁, y ₁, z ₁) and B (x ₂, y ₂, z ₂), the planimetric coordinates of query point C is (x ₀, y ₀), put C (x ₀, y ₀, z ₀) height be:

$z_0=(z_2-z_1)*\frac{S_1}{S_2};$ Formula (13)

In formula (13), S ₁for A point and the distance of C point time; S ₂for the distance between A point and B point.

Above-mentioned steps S5 is: the thread in GPU is arranged to synchronous point, judge the calculating that just continues next step, to guarantee to read the correctness of intervisibility result until the interior all threads of same thread block all complete every some intervisibility on sight line line segment.In CUDA framework, realize the setting of synchronous point by fence (barrier), call syncthreads function.

Above-mentioned steps S6 is: the each thread block in GPU is by the intervisibility judged result in parallel traversal method shared drive, if the height of all discrete points is all greater than the height of the landform of this position in sight line, judge this sight line intervisibility, otherwise judge not intervisibility of this sight line, and the intervisibility result of this sight line is kept on the video memory of GPU.

As shown in Figure 3, above-mentioned parallel traversal method completes parallel computation by tree-shaped addition: by intervisibility result value of being designated as 0, not intervisibility result value of being designated as 1.By tree-shaped addition by the summation process parallelization of the intervisibility result of all threads.If tree-shaped addition rreturn value is 0, sight line intervisibility, otherwise not intervisibility of sight line.Sight line intervisibility result is write back to shared drive, as shown in the label 4 in Fig. 2.

Above-mentioned steps S7 is: the intervisibility judged result of all sight lines that GPU is obtained writes back CPU internal memory, copies to CPU internal memory from the video memory of GPU by pci bus.

The sight line intervisibility result in shared drive is write video memory by each thread block, as shown in the label 5 in Fig. 2; Then read video memory by CPU and obtain all intervisibility results, as shown in the label 1 in Fig. 2.

Should be appreciated that the above detailed description of technical scheme of the present invention being carried out by preferred embodiment is illustrative and not restrictive.Those of ordinary skill in the art modifies reading the technical scheme that can record each embodiment on the basis of instructions of the present invention, or part technical characterictic is wherein equal to replacement; And these modifications or replacement do not make the essence of appropriate technical solution depart from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (10)

1. the intervisibility method for parallel processing based on GPU, is characterized in that, comprises the steps:

S1: build GPU programmed environment;

S2: calculate the sight line that connects observation point and impact point;

S3: write data to GPU;

S4: every some intervisibility situation on parallel computation sight line line segment;

S5: synchronous point is set;

S6: walk abreast and judge intervisibility result;

S7: sense data is to CPU.

2. the intervisibility method for parallel processing based on GPU according to claim 1, it is characterized in that, described GPU programmed environment comprises hardware environment and software environment, and wherein hardware environment comprises CPU, supports the display chip of CUDA framework and connects the pci bus of CPU and display chip; Software environment comprises C/C++ compiler and CUDA.

3. the intervisibility method for parallel processing based on GPU according to claim 1, is characterized in that, described step S2 comprises following sub-step:

S2.1: read in the position of observation point, position, the terrain data of observation point and the terrain data of impact point of impact point;

S2.2: determine by the sight line line segment of observation point and impact point on CPU;

S2.3: the position of observation point and impact point is converted to geocentric coordinate system from geographic coordinate system;

S2.4: calculate the sight line line segment between observation point and impact point under geocentric coordinate system.

4. the intervisibility method for parallel processing based on GPU according to claim 3, is characterized in that, the terrain data of observation point and impact point respectively comprises longitude, latitude and height.

5. the intervisibility method for parallel processing based on GPU according to claim 1, is characterized in that, described step S3 comprises following sub-step:

S3.1: the video memory that all data is written to GPU from the internal memory of CPU;

S3.2: large and that remain unchanged data volume terrain data is put into texture cache from video memory and accelerate to read;

S3.3: constant buffer memory is put into in frequent observation station of accessing and impact point position in calculating and accelerate to read.

6. the intervisibility method for parallel processing based on GPU according to claim 1, is characterized in that, described step S4 comprises following sub-step:

In S4.1:GPU, a corresponding sight line line segment of thread block, distributes shared drive to preserve communication data;

S4.2: the thread mean allocation in each thread block is calculated the part discrete point in sight line;

S4.3: all threads are carried out identical discrete point intervisibility judgement simultaneously.

7. the intervisibility method for parallel processing based on GPU according to claim 6, it is characterized in that, described step S4.3 is: utilize interpolation computing method to be calculated the geocentric coordinate of corresponding discrete point by the geocentric coordinate of observation point and impact point, calculated the longitude of this discrete point by the geocentric coordinate of this discrete point, latitude and height, obtain the height of the landform of this discrete point position according to the longitude of this discrete point and latitude inquiry terrain data, by relatively judging whether the height of this discrete point is greater than the height of the landform of this discrete point position, and judged result is write to shared drive.

8. the intervisibility method for parallel processing based on GPU according to claim 1, it is characterized in that, described step S5 is: the thread in GPU is arranged to synchronous point, judge until the interior all threads of same thread block all complete every some intervisibility on sight line line segment the calculating that just continues next step.

9. the intervisibility method for parallel processing based on GPU according to claim 1, it is characterized in that, described step S6 is: the each thread block in GPU is by the intervisibility judged result in parallel traversal method shared drive, if the height of all discrete points is all greater than the height of the landform of this position in sight line, judge this sight line intervisibility, otherwise judge not intervisibility of this sight line, and the intervisibility result of this sight line is kept on the video memory of GPU.

10. the intervisibility method for parallel processing based on GPU according to claim 1, is characterized in that, described step S7 is: the intervisibility judged result of all sight lines that GPU is obtained writes back CPU internal memory, copies to CPU internal memory from the video memory of GPU by pci bus.

Images